The quality of a weld joint is adversely affected by disturbances and parameter variations that occur during the welding process. In order to improve weld quality several researchers have developed automated welding systems which regulate the puddle geometry. In the past such systems have been designed using PI or PID controllers with little or no effort being made to model the dynamics of the puddle geometry. In addition arc welding processes are frequently modeled as single-input single-output systems. In this paper both single-input single-output (SISO) and multi-input multi-output (MIMO) models of the puddle geomtry are found using off-line identification techniques. Optimal linear quadratic (LQ) control methods are used to design two separate control systems. The first regulates puddle width while the second jointly regulates puddle width and area.